Activity › Forums › Blackmagic Design › RAID for Decklink HD
-
RAID for Decklink HD
Posted by Austin Wallender on February 3, 2006 at 10:57 pmHi everyone,
We’re upgrading the RAID on one of our systems handle HD digitizing on our Decklink HD, and we’re looking at raids from Medea, Huge, as well as SATA raids from promax and thedrgroup…
What are most people using for storage these days? It seems like many of you are building your own SATA RAIDs, but are there any VAR solutions that you would recommmend? I’ve looked at the BMD site, but I’d love some real world reports…
Thanks in advance,
-ausitn
Bill Buchanan replied 20 years, 3 months ago 4 Members · 7 Replies -
7 Replies
-
Bill Buchanan
February 4, 2006 at 1:17 amCheck out Broadcom’s RAIDCore SATA controllers. I’ve been using 2 of their 4852 cards to connect/control 14 300gb Seagates setup as 2 RAID-5 arrays. Using BMD disk speed utility, I’m getting about 380mb/s Read and about 285 Write. While my work is still SD, I don’t know whether those speeds and the overhead that comes with them will work with your HD material.
Bill Buchanan
Buchanan Film Co. -
Mamba_admin
February 4, 2006 at 8:56 amI’m achieving Read/Write speeds of 480MB/s.
My config is as follows:
10 Seagate 200GB SATA Drives – 8 Connected to a Sonnet Tempo-X SATA 4+4 + 2 on Apple Internal Bus.
I’ve created a RAID level 0 volume with these.
My System is located on a 200GB WiebeTech Fire800 External.Throughput is plenty good enough to cut HD with.
-
Bill Buchanan
February 4, 2006 at 4:08 pmWhoa! You folks that choose to run RAID-0 are either super confident souls, have never had a drive in your array fail or enjoy recapturing footage. Is the less-than-dramatic difference in Read/Write speeds (and overhead) between RAID-0 and RAID-5 absolutely necessary? If it is, perhaps you have no choice. I hope you have a way to backup all your vid files (and have done so), or you really don’t have all that much on your array(s)at a given time, so that a drive failure won’t cause your mouth to go dry and your life/career to flash before your eyes.
If possible, setting up a RAID-5 array to test whether or not it is fast enough for your kind of work would be time well spent. Living on the edge is fine, until Murphy shows up.
Bill Buchanan
Buchanan Film Co. -
Mamba_admin
February 4, 2006 at 6:06 pmActually, the speed difference is significant. For instance, on a 14 disk 3.4 TB XServe RAID(RAID 50) I measured throughput at 365MB/s, whereas with my 10 disk 1.8TB array(RAID 0) I measured throughput at 480MB/s.
You may remark that 365MB/s is way fast enough for HD, but bear in mind that performance decreases markedly after the drive becomes over half full.However, you’re 100% right: A RAID array without fault tolerance is foolhardy, to say the least. If the OEM option is out of one’s price range, I’d recommend two RAID 5 PCI 8 port cards with ‘hot swap’ and ‘hot spare’ capability. Populate both cards fully, stripe each with RAID 5 including one hot spare on each. Then, either in MacOSX Disk utility, or in the cards’ RAID Admin utility(if it’s supported), stripe across both arrays with RAID 0, effectively resulting in a RAID 50 array. Performance should be more than satisfying.
Bear in mind that, at present, the above is not yet possible in the newest(PCIe) G5’s. -
Casey Mershon
February 4, 2006 at 8:15 pmHi Bill,
Could you please elaborate a bit about your PC. I am wondering what MOBO you are using, and what type of PCI slots it has.
I am also using the 4852 with 8 SATA drives, but my speeds are nowhere near yours.
I am using a Supermicro P4SCT+II MOBO (Single P4 3.4 Ghz) with 64bit, 66 Mhz PCI slots.
Thanks for your reply.
Casey Mershon
-
Bill Buchanan
February 4, 2006 at 10:23 pmCasey:
Here’s the deal…
MB: SM X5DA8
Relevant PCI slots include 2 100mhz and 1 133mhz, etc. Look up Supermicro’s web page on this MB.Vid Storage/RAID setup:
14 300gb Seagate SATA HDDs
2 4852 controllers, in 100mhz slots. 8 drives connected to one controller and 6 connected to the other. 2 RAID-5 arrays (each 1.9TBs) across both the controllers, including all the drives.
The speed difference you are concerned about is probably because all 14 of my drives are part of the “Volume” being tested.
One issue I ran into when I first setup the 4852s was that one (or two, can’t remember) of the HDDs individual throughput speeds was way down compared to all the rest (using iometer). Individual speeds using iometer should run from about 50 to 63mb/s Read and Write. If you have any HDDs testing below that, you need to exchange it/them. One slow drive will slow down the entire array. In my case, the faulty drive ran about 20mb/s–definitely faulty. If your speed tests are producing widely varying results and/or very low numbers, you might want to run iometer on each drive. It’s a hassle, but the only way I know of to find the bad guy(s), if any.
Bill Buchanan
Buchanan Film Co. -
Bill Buchanan
February 4, 2006 at 10:31 pmCasey:
I was wrong. Just remembered something. If you open RAIDCore Management Console, pull down one of the menus (don’t have that computer turned on presently) and find the SMART check. With it you can see what’s going on with each drive in your array and where individual throughput speeds are displayed. That way you don’t need to hassle with iometer.
Bill
Reply to this Discussion! Login or Sign Up